AQL - How to count samples with defects for each defect class

Chris64

Registered
Steve, I agree with you, all that must be done and has been done AHEAD OF TIME.

So, let's put everything together now... AHEAD OF TIME:

My Sampling plan is based on ISO2859 and defines:
a. Acceptance Criteria: 2,3 (Acc,Rej) (=Given as result of LOT size and AQL)
b. Defect Class: Minor
c. Number of Characteristics to check: 3
d. It takes 1 characteristic - no matter which one - for a unit to count as defective.

And here are the results of the inspection:

Sample 1 is defective (fails on Characteristic 1)
Sample 2 is defective (fails on Characteristic 1)
Sample 3 is defective (fails on Characteristic 2)

Now I need your help to count the defect units:

METHOD A: I count 3 defective units in total ==> LOT REJECTED
METHOD B: I have 2 defective units for Characteristic 1 and 1 defective unit for characteristric 2. The maximun number of defective units for all characteristics under inspection is 2 ==> LOT ACCEPTED

Which is the correct method to count the defect units in your opinion?
 

Steve Prevette

Deming Disciple
Leader
Super Moderator
Given the sampling plan is "it takes 1 characteristic - no matter which one - for a unit to count as defective", you have three units that are defective, and the lot should be rejected.
 

Chris64

Registered
Hi Steve,

this is a very interesting approach. If I got you right, I have the freedom to define my own definition of a defective unit. In that case I would choose to change my prior definition to something that will not give me so many rejections and problems with the suppliers.

Using the same data of the above example, I define a defective unit as follows:

d. It takes all characteristics for a unit to count as defective

According to this new definition, the lot of the above example shall be accepted. BINGO!

BUT... The ISO2859 defines a "nonconforming item" as an item with one or more nonconformities" (3.1.7). This definition contradicts my own definition.

Now, I know some FDA inspectors who like to dig deeper and ask questions like this one: Why did you define the defective unit in such a way, i.e. against the definition contained in the standard?

Since the ISO2859 doesn't give an answer to this question, what would be a good rationale, according to your experience?
 
Last edited:

Steve Prevette

Deming Disciple
Leader
Super Moderator
Yes, you can make all your defective units disappear with a piece of paper and a signature!

Of course, may want to look at the new Cove discussion - do morals matter any more?

The standard does require a good faith effort to define the critical, major, and minor defects. And a sudden shift to having no defects (reported) may be questioned by the customer.

Consider the Peanut Corporation of America, and the owners who are in jail now, and looks like the QA manager just got let out of jail this year.
 

Chris64

Registered
My question was raised because of your statement/idea to feel free to define the defective unit as one wishes:

"...if there are multiple things being monitored in the same sampling, then you've got to write ahead - it takes 1 Apple and 3 Oranges and 2 Cherries to count the unit as defective."

The conclusion is, that you cannot do that, you cannot define the defective unit as you wish (see ISO2589 3.7.1), because, as you recognized "...you can make all your defective units disappear with a piece of paper and a signature! "

Thank you for your help, Steve!
 

Sokat

Registered
Aren't sampling plans based on the number of defective units, not the number of defects?

Actually, this is the fundamental question. Is there a clear answer to this. If either of the characteristics does not affect the intended purpose of use, then the sample would not be considered 'defective' ?
 

Steve Prevette

Deming Disciple
Leader
Super Moderator
Actually, this is the fundamental question. Is there a clear answer to this. If either of the characteristics does not affect the intended purpose of use, then the sample would not be considered 'defective' ?

No one can answer this question except the designer of the sampling plan, the supplier, and the customer. The simplest situation is I only have one characteristic I am checking for, and if that characteristic "fails" (and I have to define what is failure) then I have a defect and the unit is defective and I apply that to the number of defective units I need to declare the lot a failure. The simplest case is the c=0 plan where this would involve one defect = one defective unit = a rejected lot.

In some cases it may be worthwhile to add the complexities that have been discussed in the thread. But without intimate knowledge of the materials, the production process, and the customers' requirements and needs, no one on this discussion board can give you a "Clear answer".
 
Top Bottom